‘The frog is boiling’: is 2025 when AI finally comes for your job?

This post was originally published on this site.

Marc Benioff, billionaire chief of the $315 billion business software giant Salesforce, is set to lay off 1,000 people and said that in 2025 he would not hire any new engineers. In another era, telling the world you’re not bringing in any new hot-shot coders would be an admission of defeat, that your company was in trouble.

But this is not any era. This is the age of artificial intelligence. Consider Google, which quietly asked 20,000 workers in its Android and Pixel operations — a tenth of its global workforce — to “voluntarily exit” amid the AI boom.

Mark Zuckerberg, meanwhile, has laid out plans to cut 5 per cent of Meta’s 72,000-strong workforce, just weeks after warning that 2025 will be the year that AI engineering “agents” will rival mid-level human coders. Zuckerberg said: “This is going to be a profound milestone and potentially one of the most important innovations in history.”

Since ChatGPT crashed onto the scene two-plus years ago, we have been hearing that soon — any day now — the machines would come for our jobs. AI would supplant humans across vast swathes of the economy; our days as productive members of society were numbered.

That tsunami has not yet materialised. But if AI developers and Silicon Valley executives are to be believed, 2025 will be the year that changes. Why?

For one, businesses have spent two years experimenting, getting comfortable with what AI can do, what it can’t, and how it can best slot into their business. In the meantime, AI capabilities have improved dramatically as rivals rush out more powerful models seemingly by the month.

“Basically, ChatGPT happened, and the frog’s been boiling for a couple of years with no one really noticing,” said Jack Clark, policy chief at Anthropic, the $40 billion AI start-up behind popular chatbot Claude. “We’re gearing up for this year to be a year where you’ll have some, ‘oh shit’ moments.”

• A Sputnik moment, but has China really outsmarted the US on AI?

Clark pointed to the rise of “agents” as one such trigger. Agents are AIs that have, effectively, been liberated from a search box on a website or app. Instead, these AIs can go out onto the internet to conduct research, book tickets and run errands on the behalf of a human or business.

Both Anthropic and OpenAI have launched such capabilities in recent weeks. Google is expected to unveil a version of its universal assistant Astra by June.

These agents are, for now, limited. Casey Newton, of tech blog Platformer, asked OpenAI’s “Operator” agent to buy him groceries, only to watch it begin filling a cart in Iowa, nearly 2,000 miles away from his home in San Francisco. And yet, he said of another task, there was “something striking about telling a computer to do something and then watch it open a browser tab, type queries, click on buttons and present you with a report”.

The clunky agents of today point the way toward a goal the industry has been talking about for years: AIs that are unleashed into the world and autonomously get things done on your behalf. George Sivulka, founder of Hebbia, a $700 million developer of AI tools for businesses, predicted that this was the thin end of the wedge. “The majority of the US GDP will be created or attributable to AI agents in the next decade,” he said. “That’s my prediction.”

It is not all bad for humans. Salesforce may have frozen hiring for software developers, but does plan to bring on 2,000 salespeople to sell its AI tools. The question is: how long before the AIs come for them as well?